A useful long-tail keyword expresses a specific need your product actually meets. Length alone does not make a query relevant or cheap. Start from product facts and observed shopping terms, then test whether the traffic can support your economics.

Build a product-fact vocabulary

List verified attributes: size, material, capacity, compatibility, use case and included components. Do not add an unsupported attribute because it appears in a popular search phrase.

For a hypothetical 500 ml stainless-steel bottle without insulation, “500 ml stainless steel bottle” describes supported facts. “Insulated hiking flask” does not. Compatibility terms need particular care: verify the supported model and avoid implying an affiliation that does not exist.

Mine three sources with different evidence strength

Source	Useful for	Limitation
Advertising search term report	Actual queries that received your ad traffic	Limited to traffic captured by those campaigns
Product questions and your verified catalog facts	Language around suitability and use	A question is not proof of search volume
Manual research or keyword tools	Candidate discovery	Estimates require validation; they are not your sales results

Group candidates by shopper intent rather than generating every possible word combination. Keep discovery hypotheses separate from terms with mature orders.

Validate before promotion

For each term, check product fit, intended campaign, existing target and relevant negatives. Compare clicks, spend, orders and sales from a consistent mature period. Amazon match types can include close variations; inspect actual search terms after launch rather than assuming exact means a literal-only query. Amazon match-type guidance.

A relevant term with no recent orders may still have incomplete reporting or too little traffic. Use an affordable test limit. A clearly incompatible term does not need spending to establish the mismatch.

Set the test economics

Hypothetical example: £35 revenue per order, 30% target ACoS and 4% expected CVR imply an affordable expected CPC of £35 × 0.30 × 0.04 = £0.42. If the observed CPC is higher, investigate whether conversion or contribution can support it. These assumptions are not category benchmarks or promised results.

Keep a discovery budget separate from established traffic. Promote a supported candidate into manual targeting when it merits a dedicated bid and review. Routing negatives are optional controls, not an automatic reward for a single order.

Keep the keyword count auditable

New inventory example: start with 12 active terms, add 15, 45 and 28 in three reviewed batches, then remove 10 duplicates or unsuitable entries. The result is 12 + 15 + 45 + 28 − 10 = 90 active terms, a net increase of 78. Added count, active count and net growth are different measures.

Term	Product_fact	Source	Intent_group	Status	Test_limit	Review_date
example_query	verified_attribute	search_term_report	assign_group	candidate	set_limit	set_date

This is a manual worksheet. Use current provider templates for uploads and verify the resulting target scope.

Review value, not just list size

Track spend, mature sales and contribution at the term or intent-group level. Report whether sales means attributed sales or total business sales. More keywords can split sparse data without increasing useful reach, so adding a fixed number each week is not an objective by itself.

Refresh earlier periods after the applicable attribution window and reporting delay. Amazon conversion definitions.

Should I use broad, phrase or exact?

Choose based on the exploration you want and the traffic you can afford. Read match types by KPI, negative keyword design and starting bid calculation for the next decisions.

Related Arctavia resources

Use these supporting pages to compare Amazon PPC operating models and implementation choices.

• Amazon Ad Rejections: Check Claims, Images and Destinations
• Amazon Ads ACoS Suddenly Increased: What to Check First
• Amazon Ads ML Bid Review: Guardrails Before Aggression
• Amazon PPC Automation: Data, Decisions and Operating Controls
• Amazon PPC Automation Migration Checklist
• Amazon PPC Automation vs Agency: Which Operating Model Fits Better?